Semi-automatic teeth segmentation in 3D models of dental casts using a hybrid methodology - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 369

4.3 Per-Tooth Separation Using Min-Cut
As mentioned in Section 3.3 , this algorithm requires human interaction to introduce the virtual
node called source ( t ) to segment an object of interest. We can see a clear example in Figure 8 .

FIGURE 8 Segmentation applying the Min-Cut. (a) Selection of the interest point. (b) Seg-
mentation of the selected tooth.

Min-Cut requires human interaction in order to set some important parameters such as the
radius. For this reason, we designed a variant of the methodology used by Tamayo-Quintero
and Gómez-Mendoza [ 19 ] to search the virtual nodes automatically—landmarks—i.e., a point
in the centers of each object—in the ideal case—in this case each tooth, by means of NARF.
The input parameters are shown in Table 2 .

Table 2
Input Parameters for Min-Cut Segmentation

Interest Point ( x , y , z ) σ SmoothCost Radius

20,0,18 0.5 0.6 15

4.4 Semi-Automatic Segmentation (Hybrid Technique)
The results obtained by applying the hybrid technique are shown in Figures 9 and 10 . The stop
criteria are mentioned in the previous techniques and the input parameter are shown in Table
3 .

FIGURE 9 Semi-automatic segmentation proposal.

Table 3
Input Parameters Used in the Proposed Methodology

c th θ th σ SC Radius ag ss N

1 5 0.5 0.6 15 0.02 20 12

In Figure 9 , we designed a variant and this methodology is based in the hybridization of
three algorithms: Region Growing, NARF, and Min-Cut. The first step is separated the gum
from teeth using the region growing method then apply NARF and subsequently, each land-
mark is used as source in order to apply the Min-Cut. Finally, the segmentation is composed